|Publication number||US7853911 B1|
|Application number||US 11/267,655|
|Publication date||Dec 14, 2010|
|Filing date||Nov 4, 2005|
|Priority date||Nov 4, 2005|
|Also published as||US8572530|
|Publication number||11267655, 267655, US 7853911 B1, US 7853911B1, US-B1-7853911, US7853911 B1, US7853911B1|
|Inventors||Ryan Fung, Vaughn Betz, David Karchmer|
|Original Assignee||Altera Corporation|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (23), Non-Patent Citations (10), Referenced by (6), Classifications (8), Legal Events (2)|
|External Links: USPTO, USPTO Assignment, Espacenet|
The present invention relates to logic design automation. More specifically, the present invention relates to a method and apparatus for performing path-level skew optimization in a logic design tool and analyzing path-level skew in a logic design.
To communicate timing performance targets to logic design tools (or electronic design automation, EDA, tools) for field programmable logic devices (FPGAs), designers specify timing constraints such as, for example, clock period constraints, IO TSETUP requirements, and IO THOLD requirements. Based on these timing constraints, EDA tools attempt to generate design implementations which satisfy user performance targets. EDA tools also report whether the specified timing constraints have been satisfied so the designer can take steps, if necessary, to try to improve the design implementation. The designer may, for example, change EDA tool settings or impose constraints, such as placement and routing constraints, on the EDA tool.
Clock period timing constraints may be specified as performance targets. The minimum clock period for a given register-to-register path in a design is primarily a function of three delays: the maximum path delay between the two registers, the maximum path delay from the clock to the source register, and the minimum path delay from the clock to the destination register. The difference between the last two components is typically referred to as the clock skew for the two respective registers. Minimizing the maximum path delay between the two registers to minimize the clock period is the focus of most EDA optimization. EDA tools often ignore clock skew between registers during optimization, because clock signals are typically distributed on low-skew routing resources or networks. Low-skew routing resources ensure that clocks will be distributed with low “predictable” skew to allow optimization tools to focus on register-to-register delays when attempting to satisfy clock period constraints. However, when designs have more clocks than the number of available low-skew networks, some clocks need to be distributed using regular resources that are not specifically designed to be low-skew and this motivates low-skew optimization techniques and analysis.
Some inter-chip board-level communication standards rely on low-skew transfers where the goal is to have signals arrive at approximately the same time regardless of the absolute time it takes the signals to travel from source to destination. To support design implementation for these standards, low-skew optimization techniques and analysis methods are beneficial.
There are other instances, when low-skew (or zero-skew) is not of interest and so dedicated resources can not often be employed. Instead the designer would like to enforce a particular skew schedule and, in those cases, path-level low-skew optimization and analysis techniques that address skew schedules are of interest.
For the purposes of this application, low-skew or zero-skew requirements (or constraints) will be referred to as simple skew requirements (or constraints). When skew schedules are involved, those respective constraints will be referred to as complicated skew constraints. Complicated skew constraints have two types of skew schedules: source skew schedules and destination skew schedules. A source skew schedule specifies how much slower paths from different sources should be with respect to one another, for any given destination. For example, a source skew schedule may be that all paths starting on source node A are X ps slower than all paths starting on source node B, for any given destination node. The reason that the schedule only applies for any arbitrary destination node, rather than across all destination nodes, is because source skew schedules may be used with destination skew schedules. A destination skew schedule specifies how much slower the paths ending at different destinations should be with respect to one another, for any given source. For example, a destination skew schedule may be that all paths ending on destination node C are Y ps slower than all paths ending on destination node D, for any given source node. Simple skew constraints are a subset of complicated skew constraints with zero source and zero destination skew schedules, which is equivalent to no source and destination skew schedules (with source and destination skew schedules that specify that the delays of all paths emerging from sources should be equal and the delays of all paths ending at destinations should be equal). It should be noted that a source skew schedule can apply across all destination nodes rather than only for any arbitrary destination node if a zero destination skew schedule is specified. Similarly, it should be noted that a destination skew schedule can apply across all source nodes rather than only for any arbitrary source node if a zero source skew schedule is specified.
Before automatic path-level skew optimization techniques, designers in the past had to manually repair design implementations to satisfy skew constraints. This often required that the user manually insert logic, adjust the placement, and/or routing of the design, or redesign the system to accommodate the skew present in the design implementation.
Thus, what is needed is an efficient method and apparatus for satisfying path-level skew constraints. Also needed are path-level skew analysis techniques to help measure how well skew constraints have been satisfied.
According to an embodiment of the present invention, path-level skew is managed for a system. Connection-level skew slacks are identified for a system by first identifying a maximum path delay for each skew domain (for one or more skew constraints). For each skew domain, the maximum path delay and a tolerable maximum skew are used to determine absolute maximum and minimum path-level target delays for the skew domain. Long-path and short-path connection-level skew slacks are generated from a set of connection delays and the absolute maximum and minimum path-level target delays for each skew domain. The long-path and short-path connection-level skew slacks may be used to generate minimum and maximum connection-level delay budgets, for the overall design, to guide an optimization procedure to satisfy the one or more skew constraints. Techniques for analyzing and reporting path-level skew are also discussed so that designers can receive feedback as to how well a design implementation satisfies skew constraints.
The features and advantages of the present invention are illustrated by way of example and are by no means intended to limit the scope of the present invention to the particular embodiments shown:
The target device 100 includes a plurality of logic-array blocks (LABs). Each LAB is formed from 10 LEs, LE carry chains, LAB control signals, LUT chain, and register chain connection lines. LUT chain connections transfer the output of one LE's LUT to the adjacent LE for fast sequential LUT connections within the same LAB. Register chain connection lines transfer the output of one LE's register to the adjacent LE's register within a LAB. LABs are grouped into rows and columns across the target device 100. A first column of LABs is shown as 110 and a second column of LABs is shown as 111.
The target device 100 includes memory blocks (not shown). The memory blocks may be, for example, dual-port random access memory (RAM) blocks that provide dedicated true dual-port, simple dual-port, or single port memory up to various bits wide at up to various frequencies. The memory blocks may be grouped into columns across the target device in between selected LABs or located individually or in pairs within the target device 100.
The target device 100 includes digital signal processing (DSP) blocks (not shown). The DSP blocks may be used to implement multipliers of various configurations with add or subtract features. The DSP blocks include shift registers, multipliers, adders, and accumulators. The DSP blocks may be grouped into columns across the target device 100.
The target device 100 includes a plurality of input/output elements (IOEs) (not shown). Each IOE feeds an I/O pin (not shown) on the target device 100. The IOEs are located at the end of LAB rows and columns around the periphery of the target device 100. Each IOE includes a bidirectional I/O buffer and a plurality of registers for registering input, output, and output-enable signals. When used with dedicated clocks, the registers provide performance and interface support with external memory devices, for example.
The target device 100 includes LAB local interconnect lines 120-121 that transfer signals between LEs in the same LAB. The LAB local interconnect lines are driven by column and row interconnects and LE outputs within the same LAB. Neighboring LABs, memory blocks, IOEs, or DSP blocks may also drive the LAB local interconnect lines 120-121 through direct link connections.
The target device 100 also includes a plurality of row interconnect lines (“H-type wires”) 130 that span fixed distances. Dedicated row interconnect lines 130, that include H4 131, H8 132, and H24 133 interconnects, route signals to and from LABs, DSP blocks, IOEs, and memory blocks within the same row. The H4 131, H8 132, and H24 133 interconnects span a distance of up to four, eight, and twenty-four LABs respectively, and are used for fast row connections in a four-LAB, eight-LAB, and twenty-four-LAB region. The row interconnects 130 may drive and be driven by LABs, DSP blocks, RAM blocks, and horizontal IOEs.
The target device 100 also includes a plurality of column interconnect lines (“V-type wires”) 140 that operate similarly to the row interconnect lines 130. The column interconnect lines 140 vertically routes signals to and from LABs, memory blocks, DSP blocks, and IOEs. Each column of LABs is served by a dedicated column interconnect, which vertically routes signals to and from LABs, memory blocks, DSP blocks, and IOEs. These column interconnect lines 140 include V4 141, V8 142, and V16 143 interconnects that traverse a distance of four, eight, and sixteen blocks respectively, in a vertical direction.
Block 210 represents a synthesis unit. The synthesis unit 210 generates a logic design of a system to be implemented by the target device 100 (shown in
Block 220 represents a placement unit 220. The placement unit 220 fits the system on the target device 100 by determining which resources on the target device 100 are to be used for specific functional blocks and registers. According to an embodiment of the system designer 200, the placement unit 220 first determines how to implement portions of the optimized logic design in clusters. Clusters may represent a subset of the components on the target device 100 such as, for example, a LAB having 10 LEs. In this embodiment, after portions of the optimized logic design are implemented in clusters, the clusters may be placed by assigning the clusters to specific LABs on the target device 100. Following the placement of the clusters, routing interconnections between the LEs may be performed. The placement unit 220 may utilize a cost function in order to determine a good assignment of resources on the target device 100.
Block 230 represents a routing unit 230. The routing unit 230 determines the routing resources on the target device 100 to use to provide interconnection between the functional blocks and registers on the target device 100.
Block 240 represents a path-level skew optimization unit 240. According to an embodiment of the system designer 200, the path-level skew optimization unit 240 computes long-path and short-path connection-level skew slacks for a skew domain. The long-path and short-path connection-level skew slacks may be computed from one or more skew constraints, each constraint specifying a tolerable maximum skew. The one or more skew constraints may include one or more simple skew constraints where identified paths are targeted to have zero skew with respect to each other. Alternatively, the one or more skew constraints may include complicated skew constraints where a skew schedule specifies the relative skews between paths starting at different source nodes for any particular destination node and/or the relative skews between paths ending at different destination nodes for any particular source node. The long-path and short-path connection-level skew slacks may be used by the path-level skew optimization unit 240 to help generate minimum and maximum connection delay budgets to guide an optimization procedure used by the synthesis unit 210, the placement unit 220, and/or the routing unit 230.
Path-level skew optimization is more general than net-level skew optimization and, hence, may be more challenging to perform. For example, a net-level zero skew constraint simply specifies that all the connections of the net should have the same delay. With a path-level skew constraint, it is not apparent what the various connections delays should be (in relation to one another) to satisfy the constraint. Consequently, performing path-level skew optimization may be more challenging than performing net-level skew optimization. Since a path is defined as a series of one or more connections, path-level skew optimization is a superset of net-level skew optimization. The connections of a net, over which a net-level skew constraint is applied, can be thought of as a set of paths.
The synthesis unit 210 may utilize a synthesis strategy that adds or removes levels of logic, uses slower or faster variants of a functional block, technology maps into faster or slower standard logic structures (such as carry chains), and/or uses faster or slower logic inputs based on the minimum and maximum delay budgets. The placement unit 220 may utilize a placement strategy that places functional blocks so they can (or are forced to) use slower/faster dedicated routing resources, and/or places functional blocks at appropriate distances from other functional blocks, based on the minimum and maximum delay budgets. The routing unit 230 may utilize a routing strategy that requires more or less routing resources, slower or faster routing resources, and delay chains (or additional buffers) based on the minimum and maximum delay budgets.
The path-level skew optimization unit 300 includes a delay estimation unit 320. According to an embodiment of the path-level skew optimization unit 300, the delay estimation unit 320 generates minimum achievable delay estimates for connections in a logic design. The minimum achievable delay estimates are estimates of minimum delays that could realistically be achieved in a final design. According to one embodiment, the minimum achievable delay estimates may be generated using information from components such as a synthesis unit, placement unit, and/or a routing unit. Every skew constraint is associated with a set of paths (a skew domain). For each of those skew domains, the delay estimation unit 320 also identifies the delay of a maximum delay path (M), assuming connection delays equal to the minimum achievable delay estimates.
According to an embodiment of the present invention, the delay estimation unit 320 identifies a maximum delay path for each skew domain subject to a simple skew constraint by performing an analysis on a timing diagram.
According to an embodiment of the present invention, the delay estimation unit 320 identifies a maximum delay path for a skew domain subject to a complicated skew constraint by modifying the timing diagram to reflect a skew schedule and performing an analysis on the modified timing diagram. Instead of a simple skew constraint (which tries to equalize path delays within a certain tolerance), the skew domain illustrated in
Referring back to
The path-level skew optimization unit 300 includes a timing analysis unit 340. The timing analysis unit 340 generates connection-level long-path and short-path slacks from the maximum (TMAX) and minimum (TMIN) target delays. These connection-level slacks are referred to as connection-level skew slacks because if a design is optimized according to these long-path and short-path connection-level slacks, the design will be optimized to satisfy the corresponding skew constraints. According to an embodiment of the path-level skew optimization unit 300, the timing analysis unit 340 computes long-path connection-level skew slacks using a long-path timing analysis based on a set of connection delays and the maximum target delay for each skew domain. The timing analysis unit 340 computes short-path connection-level skew slacks using a short path timing analysis based on a set of connection delays and the minimum target delay for each skew domain. In some embodiments, the long-path and short-path connection-level skew slacks may be conservatively computed (using known techniques) from a set of connection delays where each connection has a minimum and a maximum delay bound to represent intra-chip delay modeling uncertainty and/or delay variation within the system (differences between rising versus falling signals, coupling between signal wires, etc.). In some of those embodiments, a technique called common-path pessimism removal can also be used to remove possible over-conservatism when computing long-path and short-path connection-level slacks when delay variation between portions of paths can be correlated. By using conservative long-path and short-path timing analysis approaches along these lines to produce slacks, etc., both path-level skew optimization and analysis techniques (which will be discussed in more detail later) can be made to account for delay variation and uncertainty. Another type of delay variation is inter-chip delay variation. Delay variation between chips may occur because of manufacturing (process) variations and operating condition (temperature and voltage) fluctuations. Coarse delay differences may arise because of inter-chip delay variation so design optimization and analysis needs to consider this. To model coarse delay differences, delay bounds are typically not used like they are for intra-chip delay variation because that would result in over-conservative analysis and optimization. Consequently, process and operating condition corners are introduced to model different “points” in the delay “space”. By optimizing a design to meet timing at each of those timing corners, and by analyzing to verify timing is met at each of those timing corners, a design can be implemented to operate robustly in spite of process and operating condition variation. Techniques that permit timing optimization for multiple timing corners may involve using conservative scaling factors to map delays from one corner to another, in order to perform analyses at multiple corners, and may involve mapping back and amalgamating slack information at a single (primary) corner so that standard single-corner optimization techniques can be used to optimize timing for multiple corners. These same techniques can be applied to optimize skew for multiple timing corners. In particular, when figuring out the maximum delay path in a skew domain with minimum achievable delay estimates, a maximum path delay would be determined at each of the N timing corners using delay mapping techniques followed by the appropriate analyses, and the N maximum path delays can be used to compute N minimum target delays and N maximum target delays (one pair for each corner). In some embodiments, the target delays may be chosen so they “correspond” as much as possible so that, at least, some design implementations can satisfy the chosen target delays at all corners. Basically, the minimum and maximum target delay at each corner determines a timing window for that corner. The overall timing window across all corners is a function of the timing windows at each of the corners and the delay differences between corners and it is advantageous to maximize the timing window across all corners to give the optimization algorithm the most flexibility to satisfy timing. Once the target delays are chosen, they can be used to guide analyses at those respective corners and the long-path and short-path skew slacks produced can be mapped back to the primary corner and conservatively amalgamated to guide optimization as if optimization was only being done for a single timing corner. That is, single-corner optimization techniques using the amalgamated skew slack information at the primary corner will implicitly optimize skew at all timing corners.
The path-level skew optimization unit 300 includes a slack allocation unit 350. The slack allocation unit 350 generates minimum and maximum connection delay budgets from the long-path and short path connection-level skew slacks. It should be noted that after producing connection-level slacks, a connection may have several connection-level slacks associated with it, for the various timing domains (skew and otherwise) it overlaps. Consequently, an overall long-path slack and an overall short-path slack can be assigned to each connection by conservatively picking the smallest corresponding slack value for each connection. These overall long-path and short-path connection slacks can be used to help generate the minimum and maximum connection delay budgets. According to an embodiment of the path-level skew optimization unit 300, the slack allocation unit 350 iteratively allocates the connection-level slacks by calling several timing analyses and conservatively allocating the slack revealed by each timing analysis. The slack allocation unit 350 may be implemented using slack allocation unit 300 and the slack allocation techniques shown in FIGS. 3, 6, and 7 in U.S. application Ser. No. 10/774,883 filed on Feb. 9, 2004 and entitled “Method and Apparatus for Utilizing Long-Path and Short-Path Timing Constraints in an Electronic-Design-Automation Tool”, herein incorporated by reference, or by any other appropriate circuitry or technique.
The path-level skew optimization unit 300 includes an optimization unit 360. The optimization unit 360 utilizes the minimum and maximum connection delay budgets to generate a strategy to satisfy the skew constraints. Since the connection level slacks are skew slacks, the strategy from the optimization unit 360 will permit the satisfaction of the respective skew constraints. It should be appreciated that the optimization unit 360 may alternatively reside in whole or in part in a synthesis unit, placement unit, and/or routing unit.
At 602, the mapped logical system design is placed. Placement works on the optimized technology-mapped netlist to produce a placement for each of the functional blocks. According to an embodiment of the present invention, placement includes fitting the system on the target device by determining which resources on the target device are to be used for specific logic elements and functional blocks.
At 603, it is determined which routing resources should be used to connect the functional blocks in the system. According to an embodiment of the present invention, routing may be performed in response to routing constraints provided.
At 604, path-level skew optimization is performed. According to an embodiment of the present invention, minimum and maximum connection delay budgets are generated in response to one or more skew constraints (that each specify a set of paths, an optional skew schedule, and a tolerable maximum skew or tolerable maximum deviation permitted from the skew schedule), other timing constraints (short-path and long-path), and minimum achievable delay estimates computed from data received by one or more of the synthesis, placement, and routing procedures 601, 602, and 603. A strategy for satisfying the one or more skew constraints may be generated in response to the minimum and maximum connection delay budgets or alternatively, the minimum and maximum connection delay budgets may be returned to one or more of the synthesis, placement, and routing procedures 601, 602, and 603 to generate a strategy for satisfying the one or more skew constraints (along with other timing constraints).
At 605, an assembly procedure is performed. The assembly procedure involves creating a data file that includes information determined by the compilation procedure described by 601-604. The data file may be a bit stream that may be used to program a target device.
At 702, absolute maximum and minimum target delays for the skew domain are generated. According to an embodiment of the present invention, the maximum target delay is chosen to be M<=TMAX<=M+S, and the minimum target delay is chosen to be TMIN=TMAX−S. It should be appreciated that the maximum and minimum target delays may be generated independently of the maximum path delay.
At 703, timing analysis is performed. According to an embodiment of the present invention, for each skew domain, a long-path timing analysis is performed on the aforementioned timing graph to generate long-path connection-level skew slacks; the timing graph is annotated with a set of connection delays and the analysis is based on a constraint equal to the maximum target delay determined. Also, for each skew domain, a short-path timing analysis is performed on the aforementioned timing graph to generate short-path connection-level skew slacks; the timing graph is annotated with a set of connection delays and the analysis is based on a constraint equal to the minimum target delay determined.
At 704, slack allocation is performed. According to an embodiment of the present invention, slack allocation is iteratively performed considering long-path and short-path connection-level skew slacks (generated by 703) to generate minimum and maximum connection delay budgets.
According to an embodiment of the present invention, a timing graph may be used to help compute the maximum path delay in a skew domain (701) and the short-path and long-path connection-level skew slacks (703). For simple skew constraints, no timing graph modifications need be performed when proceeding with the phases in
At 802; it is determined whether a skew constraint specifies relative delays between paths starting at different source nodes (source skew schedule). If a source skew schedule is detected, control proceeds to 803. If a source skew schedule is not detected, control proceeds to 804.
At 803, for each appropriate source node discussed in a source skew schedule, a delay is introduced immediately after the source node based on the skew schedule. According to an embodiment of the present invention, the delay introduced is the negative of the corresponding delay value in the skew schedule. An example is illustrated in
At 804, it is determined whether a skew constraint specifies relative delays between paths ending at different destination nodes (destination skew schedule). If a destination skew schedule is detected, control proceeds to 805. If a destination skew schedule is not detected, control terminates the procedure at 806.
At 805, for each appropriate destination node discussed in a destination skew schedule, a delay is introduced immediately before the destination node based on the skew schedule. According to an embodiment of the present invention, the delay introduced is the negative of the corresponding delay value in the skew schedule. This is similar to what is performed for a source skew schedule—which was already discussed. An example is illustrated in
The timing diagram shown in
At 1102, for a simple skew constraint, the skew of the paths in the skew domain is determined. This may be achieved by taking the difference of the maximum path delay and the minimum path delay identified at 1101. For a complicated skew constraint, this difference is the maximum deviation from the respective skew schedule.
At 1103, for simple skew constraints, the skew from 1102 is compared with a maximum tolerable skew (S) to determine the skew-domain worst slack (slack=S−skew). For complicated skew constraints, with an arbitrary skew schedule, the maximum deviation from the respective skew schedule from 1102 is compared with the maximum tolerable deviation from the skew schedule (S) to determine the skew-domain worst slack (slack=S−(maximum deviation from schedule)).
At 1104, results of the analysis are reported. According to an embodiment of the present invention, the results may be leveraged by a designer when manually repairing a design implementation to satisfy its respective constraints and/or re-designing the system to accommodate the skew present in the design implementation.
According to an alternate embodiment of the present invention, the N paths with the largest path delays and the N paths with the N smallest path delays in a skew domain (based on a timing graph for a simple skew constraint and a modified timing graph for a complicated skew constraint, whichever is appropriate) may be determined using known efficient techniques, where N may be any number. This information may also be reported to aid a designer. In some embodiments, the efficient techniques may not actually identify the N largest path delays because certain approximations are made for efficiency reasons—for example, only C paths per destination may be identified. However, in practice, such approximations are independent of the this discussion since the paths identified are usually of practical interest, and, if there are “approximations” in the paths identified, the resulting skew reports will be approximate along the same respective lines. Considering this, in another embodiment, from those N largest delay paths and N smallest delay paths, N2 pairs of paths with their corresponding skew values (or deviations) may be conceptually identified. From those N2 pairs of paths, the N pairs of paths with the largest skew values or deviations may be reported to the designer. The N pairs of paths with the largest skew values may be determined by sorting the N2 pairs according to their corresponding skew values (or deviations), in N2×log(N) time and N2 space, or by using the following procedure that runs in N×log(N) time and N space. A data structure (A) can be created to associate each of the N smallest delay paths with one of the N largest delay paths. In one embodiment, this data structure can be an array indexed by the ID of the smallest delay path, and storing values that correspond to the IDs of the longest delay paths. Initially, each of the N smallest delay paths will be associated with the largest delay path. The corresponding N starting skew values or deviations can be computed and the smallest delay paths, along with their corresponding skew values, can be inserted into a heap (H) arranged according to skew values. The heap data structure allows log(N)-time retrieval of the largest skew value inserted in the heap and the corresponding smallest delay path. After the formation of the heap, the algorithm begins by removing the largest skew value from the heap, recording the corresponding smallest delay path and the corresponding largest delay path (the corresponding largest delay path is found by looking at the appropriate entry in A); the corresponding pair can be recorded as the pairing with the worst skew (or deviation). After that, the respective entry in A can be updated to point to the next largest delay path, and the corresponding largest skew value for that new pairing can be computed and inserted in the heap in log(N)-time. Then the process can be repeated until the N worst pairs of paths (with respect to skew) are determined.
The techniques above have been described with reference to designing a field programmable gate array. It should be appreciated that the techniques (for synthesis, placement, routing, etc.) may be used in any CAD tool for the creation/processing/optimization/implementation of any logic design, such as that encountered in the creation of application specific integrated circuits (ASICs), etc.
Embodiments of the present invention may be provided as a computer program product, or software, that may include an article of manufacture on a machine accessible or machine readable medium having instructions. The instructions on the machine accessible or machine readable medium may be used to program a computer system or other electronic device. The machine-readable medium may include, but is not limited to, floppy diskettes, optical disks, CD-ROMs, and magneto-optical disks or other type of media/machine-readable medium suitable for storing or transmitting electronic instructions. The techniques described herein are not limited to any particular software configuration. They may find applicability in any computing or processing environment. The terms “machine accessible medium” or “machine readable medium” used herein shall include any medium that is capable of storing, encoding, or transmitting a sequence of instructions for execution by the machine and that cause the machine to perform any one of the methods described herein. Furthermore, it is common in the art to speak of software, in one form or another (e.g., program, procedure, process, application, module, unit, logic, and so on) as taking an action or causing a result. Such expressions are merely a shorthand way of stating that the execution of the software by a processing system causes the processor to perform an action to produce a result.
In the foregoing specification, embodiments of the invention have been described with reference to specific exemplary embodiments thereof. It will, however, be evident that various modifications and changes may be made thereto without departing from the broader spirit and scope of the embodiments of the invention. The specification and drawings are, accordingly, to be regarded in an illustrative rather than restrictive sense.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US6218864 *||Aug 10, 1999||Apr 17, 2001||Xilinx, Inc.||Structure and method for generating a clock enable signal in a PLD|
|US6466008 *||Oct 6, 2000||Oct 15, 2002||Hewlett-Packard Company||Method for matching the lengths of signal traces|
|US6550045 *||Nov 20, 2001||Apr 15, 2003||Lsi Logic Corporation||Changing clock delays in an integrated circuit for skew optimization|
|US6744285 *||Aug 8, 2002||Jun 1, 2004||Agilent Technologies, Inc.||Method and apparatus for synchronously transferring data across multiple clock domains|
|US6782519 *||Aug 29, 2002||Aug 24, 2004||Cadence Design Systems, Inc.||Clock tree synthesis for mixed domain clocks|
|US6910196 *||May 8, 2003||Jun 21, 2005||Intel Corporation||Clocked and non-clocked repeater insertion in a circuit design|
|US6948142 *||Jun 2, 2003||Sep 20, 2005||Lsi Logic Corporation||Intelligent engine for protection against injected crosstalk delay|
|US7002369 *||Apr 23, 2004||Feb 21, 2006||Texas Instruments Incorporated||Implementing complex clock designs in field programmable devices|
|US7047504 *||Feb 21, 2003||May 16, 2006||Fujitsu Limited||Method and program for designing semiconductor integrated circuits to optimize clock skews on plurality of clock paths|
|US7107477 *||Jan 31, 2003||Sep 12, 2006||Altera Corporation||Programmable logic devices with skewed clocking signals|
|US7178124 *||May 30, 2003||Feb 13, 2007||Golden Gate Technology, Inc.||Methods, algorithms, software, architectures and system for placing clocked components and routing timing signals in a circuit and/or layout|
|US7228517 *||Oct 27, 2004||Jun 5, 2007||Nec Electronics Corporation||Integrated circuit design method and system|
|US7254789 *||Dec 1, 2004||Aug 7, 2007||Altera Corporation||Optimizing long-path and short-path timing and accounting for manufacturing and operating condition variability|
|US7290232 *||Dec 1, 2004||Oct 30, 2007||Altera Corporation||Optimizing long-path and short-path timing and accounting for manufacturing and operating condition variability|
|US7296246 *||Nov 5, 2003||Nov 13, 2007||Cadence Design Systems, Inc.||Multi-domain clock skew scheduling|
|US7302657 *||Sep 26, 2002||Nov 27, 2007||Telefonaktiebolaget L M Ericsson (Publ)||Optimization of the design of a synchronous digital circuit|
|US7559040 *||Mar 10, 2006||Jul 7, 2009||Cadence Design Systems, Inc.||Optimization of combinational logic synthesis through clock latency scheduling|
|US7624364 *||May 2, 2007||Nov 24, 2009||Cadence Design Systems, Inc.||Data path and placement optimization in an integrated circuit through use of sequential timing information|
|US20040123259 *||Dec 19, 2002||Jun 24, 2004||You Eileen H.||Timing analysis of latch-controlled digital circuits with detailed clock skew analysis|
|US20060080632 *||Jan 25, 2005||Apr 13, 2006||Mathstar, Inc.||Integrated circuit layout having rectilinear structure of objects|
|US20070035428 *||Jul 13, 2006||Feb 15, 2007||Mustafa Badaroglu||Method and apparatus for minimizing the influence of a digital sub-circuit on at least partially digital circuits|
|US20070204252 *||May 4, 2007||Aug 30, 2007||Furnish Geoffrey M||Methods and Systems for Placement|
|US20070234266 *||Feb 7, 2004||Oct 4, 2007||Chao-Chiang Chen||Method of optimizing IC logic performance by static timing based parasitic budgeting|
|1||*||Albrecht et al.; "Cycle time and slack optimization for VLSI-chips"; Publication Year: 1999; Computer-Aided Design, 1999. Digest of Technical Papers. 1999 IEEE/ACM International Conference on; pp. 232-238.|
|2||*||El-Amawy et al.; "On the complexity of designing optimal branch-and-combine clock networks"; Feb. 1998; Computers, IEEE Transactions on; vol. 47, Issue 2, pp. 264-269.|
|3||*||Fung et al.; "Simultaneous short-path and long-path timing optimization for FPGAs"; Nov. 7-11, 2004; Computer Aided Design, 2004. ICCAD-2004. IEEE/ACM International Conference on; pp. 838-845.|
|4||*||Harris et al.; "Statistical clock skew modeling with data delay variations"; Dec. 2001; Very Large Scale Integration (VLSI) Systems, IEEE Transactions on; vol. 9, Issue 6, pp. 888-898.|
|5||*||Kwang-Ki et al.; "Skew optimization by combining tree-based and graph-based techniques for high performance clock routing"; Oct. 26-27, 1999 ; VLSI and CAD, 1999. ICVC '99. 6th International Conference on; pp. 407-410.|
|6||*||Neves et al.; "Optimal clock skew scheduling tolerant to process variations"; Jun. 3-7, 1996; Design Automation; Conference Proceedings 1996, 33rd; pp. 623-628.|
|7||*||Ravindran et al.; "Multi-domain clock skew scheduling"; Nov. 9-13, 2003; Computer Aided Design, 2003. ICCAD-2003. International Conference on; pp. 801-808.|
|8||*||Schmid et al.; "Advanced synchronous scan test methodology for multi clock domain ASICs"; Publication Year: 1999; VLSI Test Symposium, 1999. Proceedings. 17th IEEE; pp. 106-113.|
|9||*||Singh et al.; "Constrained clock shifting for field and programmable gate arrays"; Feb. 2002; FPGA '02; Proceedings of the 2002 ACM/SIGDA tenth international synposium on Field-programmable gate arrays' ACM.|
|10||*||Yeh et al.; "Delay budgeting in sequential circuit with application on FPGA placement "; Publication Year: 2003; Design Automation Conference, 2003. Proceedings; pp. 202-207.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US8079004 *||Apr 30, 2009||Dec 13, 2011||Synopsys, Inc.||Efficient exhaustive path-based static timing analysis using a fast estimation technique|
|US8572530||Dec 6, 2010||Oct 29, 2013||Altera Corporation||Method and apparatus for performing path-level skew optimization and analysis for a logic design|
|US8769470 *||Nov 15, 2011||Jul 1, 2014||International Business Machines Corporation||Timing closure in chip design|
|US20100281445 *||Apr 30, 2009||Nov 4, 2010||Synopsys, Inc.||Efficient exhaustive path-based static timing analysis using a fast estimation technique|
|US20120137263 *||Nov 15, 2011||May 31, 2012||International Business Machines Corporation||Timing closure in chip design|
|US20150220674 *||Dec 23, 2014||Aug 6, 2015||Tabula, Inc.||Detailed Placement with Search and Repair|
|International Classification||G06F17/50, H03K19/00, G06F9/45|
|Cooperative Classification||G06F17/5054, G06F1/10|
|European Classification||G06F17/50D4, G06F1/10|
|Nov 4, 2005||AS||Assignment|
Owner name: ALTERA CORPORATION, CALIFORNIA
Effective date: 20051103
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:FUNG, RYAN;BETZ, VAUGHN;KARCHMER, DAVID;REEL/FRAME:017192/0499
|May 28, 2014||FPAY||Fee payment|
Year of fee payment: 4